AI in Adolescence: Twilight Zone or Asimov’s Robotics?

AI is growing up fast, becoming more powerful. The real question is not whether it will mature. The real question is whether we will.

And that's what Shimmy says. Hey everyone, it's Shimmy, and welcome to this week's Shimmy says, I appreciate you coming down, uh, or joining in. You're not coming.

I wish you were coming down here, but no, you're just joining in. Anyway, this week, shimmy says, I call AI in adolescence. The question is, is it a twi, is it a Twilight Zone or are we talking about Isaac Asimov's laws of robotics?

That's a good question. Let's get into it. So this week I read a, it's a blog post, but it's really an essay by Dario Am Modi, who's the CEO of philanthropic.

And that led me to a another essay that he wrote actually almost two years, a year and a half ago before that. The one that he wrote just like this week, or maybe it was last week, is called The Adolescence of Technology. And then the earlier essay is called The Machines of Loving Grace.

And let me just say this upfront, there are two of, I think, the most thought provoking, smartest things that I've read in the whole AI space in a long time. And they're really sort of in yin and yang of what AI is and what it could be. It's not hype, it's not fud or fear porn, or whatever you want to call that.

It's not marketing fluff either. It really, it causes you to think you can't just run through these things. You gotta focus in and think.

And I read them, I reread them, you know, and then I read 'em back to back and it got me thinking. And the way I always think, you know, I'm, I'm a child of my culture, and when I see these things, I immediately start thinking through movies, TV series, books, songs, culture, because sometimes I need that to make sense of big, big ideas like that. And the phrase that struck me in, in e Modi's essays was right in the title, right, the adolescence of technology.

Really, it's the adolescence of ai. And he, he makes a strong case that AI is in its adolescence. But if we truly think that AI is in its adolescence at some level, that should scare the hell out of us.

Because for anyone who's been a parent, you know what it's like having a raising adolescence. It scares that, it scare the hell out of us. But also it, if you're a parent and you know, right?

It gives you all kinds of hope. 'cause you know, deep in there, there's this amazing thing, this amazing being, and maybe that's what AI is. So let me explain to it.

It's almost like AI is a teenager and he's got, they've got the car keys. You know, when the No Emoti talks about AI being in an adolescent phase, he's really talking about the cap. Its capabilities racing ahead of its maturity, right?

That middle school kind of demon phase power growing faster than wi wisdom systems that can reason, plan, act, and influence real outcomes. While quite frankly, US society, institutions in government and governance are still trying to catch up. We're playing catch up, we're behind.

And this isn't theatrical anymore. It's happening right now. You know, when I hear an adolescence with immense power as a Modi called it, again, being who I am, I'm thinking TV's, movies.

My brain goes straight to an old Twilight Zone episode that used to scare the hell outta me. That episode was called, it's a Good Life. Now, I don't know how many of you are original Twilight Zone fans, but in that episode, it was actually Billy Mummy, who was the original Will Robinson in the original Lost in Space series, bill Murray's, uh, bill Mum Billy Mummy's, a young boy, and he has like God-like powers, he could read minds.

He knows what you're thinking. He could change reality just with a thought. He can make you disappear, disappear if he doesn't like what you're thinking in the show in Twilight Zone, he used to send people out into a cornfield in the remake of the Twilight Zone, the movie he would like cover their mouth, so they couldn't speak and stuff as a result of that.

This little town that the boy Billy Mummy's, you know, character lives in, they live in terror. Everyone smiles, everyone praises him. Everyone does everything he does.

Nobody tells him no, because if you do or you make him upset, you're out in the cornfield. Some people say, oh, Billy Mummy's evil. No, Billy's not evil.

And that episode isn't evil. It's about unchecked power without maturity. And if we think of AI being sort of omnipotent at some level and, and not having that maturity, maturity, that's a future.

None of us, none of us want. It's not because AI wakes up one morning and decides to be evil mind you, but, but it's because it gets powerful before it gets wise because it incent its incentives. Get misaligned.

Feedback loops break because the people building and deploying these systems lose control of the outcomes in their headlong rush to money. It, it, you know, this is what get pe gets people, especially on the inside, like Emoti worried about these things, and he's very clear about this risk. He doesn't minimize it, he doesn't sugarcoat it, he doesn't dance around it.

He flat out says that AI systems are becoming more powerful faster than our ability to govern them responsibly. And that gap right there is where things can go sideways in a bad way. Guys, this isn't science fiction or Twilight Zone anymore.

Here's the part that really matters. AI is no longer just a tool that sits on the sidelines. These systems are, are acting right now, acting inside real workflows.

They influence decisions, they're generating content, they automate actions, they shape markets. They're increasingly embedded into the very fabric, the machinery of society. And that's a big shift in a really short time.

We haven't caught up to it. And again, as a parent, you know, this a or if you're a child behaviorist or you know, psychologist, adolescence is the most dangerous phase of growth. As I said, if you raise kids, you know that it's smart enough to do jet damage confident enough to think they know better, but not yet mature enough to understand the consequences.

Now, let's scale that up globally. That's the concern of Modi's raising. And quite frankly, I think he's right to raise it.

I, he really made me think about it. History tells us we are terrible at slowing down when money, power and competition are involved. We ask, can we long before we ask, should we?

And that's kind of a real adolescent move, isn't it? We almost never ask what happens if we're wrong? 'cause when you're young, you don't think you could be wrong.

Of course, though, we've seen this movie before, right? We've seen it recently over the last decades. Social media, surveillance, capitalism, algorithmic engagement, all built fast, scaled faster, and then governed, laid later if it was governed at all.

So, you know, we've all joked about AI being like Skynet, right from the old movie. And I, I get it's, it's something funny to talk about, but the real danger isn't some cartoon villain in AI like Skynet here. I think the real danger is indifference.

If we don't do anything, we have optimization without context and intelligence, without values. That's the Twilight Zone version of AI's adolescence. But here's the good news.

That's only half the story because these essays really stuck with me and made me think it through. Emoti isn't just warning us, he, he's also offered us a path forward. The earlier essay, machines of Loving Grace gave us, gave us this path forward.

In that essay, emoti steps back and ask big questions. If machines are going to become more intelligent than the smartest humans, super intelligent, omnipotent, how do we make sure that intelligence is aligned with human values? Not abstract values, but real ones, messy ones, human ones.

What I appreciated is that he didn't pretend that, that this is, this is an easy thing to tackle. He doesn't pretend as that is even a single right answer. He acknowledges that humans themselves don't agree on this.

Aligning machines is not a checkbox, it's gonna be an ongoing process, but that's a critical point. Alignment itself is not a destination, it's just part of the journey. It's a discipline.

And he introduced something called the Claude Constitution. Some people may call it the Anthropic constitution. This is one of the most interesting things that I've seen recently, right?

O Modi talks about the Claude Constitution. They've recently updated. It's not some marketing gimmick, or at least I don't think it is, fooled me if it is, it's an actual attempt to define the principles and norms that should guide how AI systems behave.

That will guide supposedly how clawed from anthropic behaves. The models are trained to reason against a written set of values drawn from things like human rights frameworks and ethical principles. In its simplest terms, the system isn't just ask what is the answer?

It's also ask, is this the right answer to give and why? And look man, that's a big deal for where we are today with ai, because one of the biggest fears around AI is not that it's malicious, that that is malice. It's misaligned systems doing exactly what they were asked to do, just not what we really want them to do.

Now a constitution doesn't solve that completely, and Imodium self is very clear about it. No static set, no static set of rules ever. Could be that, but it's a hell of a lot better than letting behavior emerge random randomly from scale and data and greed alone.

And that's where my second cultural reference kicked in, right? This time, it's not a movie, I go back to books. It's Isaac Asimov.

You know, when I was a teenager, I was in Asimov in a big way. I read all, all of his sci-fi books reading about this, uh, Claude Constitution. I couldn't stop thinking and analogizing it to Isaac Asimov and his laws of robotics.

Most people, if if, well if you were any kind of sci-fi fan, even if you weren't, you may have heard about these laws of robotics. Robots can't harm humans. They must obey humans.

They must protect themselves unless that conflicts with the first two laws. And so the highest law was they can't harm humans. And if that means disobeying humans, not to harm them later, if you're a fan of Asimov, you know, there was another zero with law, a fourth law putting in there that the greater good for all of humanity outweighs the good to just the individual.

Yeah, you may say, ah, it's old stuff in a book from the forties and fifties. But what made Asimov brilliant wasn't that he thought these laws were foolproof. It was that he as an author, understood that these rules were gonna create tensions.

That scenarios would come up. Edge cases, moral dilemmas that would make terrific theater, terrific reading. His stories are all about what happens when intelligent systems follow rules perfectly.

And humans have to deal with the consequences. That's the parallel here. Philanthropics constitution, clawed constitution is not a guarantee, it's just an attempt.

It's an acknowledgement that intelligence without guardrails is dangerous. And that values need to be part of the system. Not an afterthought, but let me stop right here for a second and go back to Asimov.

They wrote a whole constitution. Claude Asimov did this in four rules. The simplicity, the sheer simplicity and elegance of asimov's rules of robotics.

And you could just substitute AI for robotics, the sheer simplicity and elegance. But that it was encompassing. Everything is something I think we strive for and we will strive for.

No one wants to read through a whole constitution. We weren't defining principles. But if I had to have, if I had to pick one thing with this constitution is maybe it tries to be too much.

It tries to put too much down in paper. It tries to anticipate too much. We need a simpler, elegant set of rules, is what I think.

And when I think about the real world for a second, that's kind of where I'm at, is that this is a good start. But we need to be, we need to boil this down to its essence. You know, the other thing to think about is look where it, remember where it's coming from, right?

Andro announced, I think it was last week, another massive funding round was a $20 billion oversubscribed round. They we're looking to only raise 10 billion. Like 10 billion is mere drop in the bucket.

They're valuing anthropic now at $350 billion. Guys, that's not a headline. That's real power open ai.

Rumor has, it is raising another round too. They're gonna be valued just under a trillion dollars when AI companies are valued as much as nation states. Governance is no longer an internal discussion within the company.

It's a societal, it's a societal humanity wide issue. And we gotta treat it as that. Give credit.

Emoti doesn't duck this at all in these essays. He talks openly about the responsibility that comes with building these awesome frontier models. What we all use now about the roles executive and boards are playing in shaping outcomes about the fact that leadership decisions matter as much as technical wants.

That honesty matters. It matters to me because I think he lets it, he comes across as real. And again, if he fooled me, shame on them.

The biggest risk in AI may not be in the models themselves. It may be the incentives that are driving the people who build these models. So guys, here's my shimmy's take on this one.

I don't know Dario or Modi personally, I didn't really wasn't aware of him much before. I read these two essays. I'm not inside anthropic open AI or any of 'em.

But based on these uh, essays, I gotta tell you the truth, I came away really impressed with him and, and the thought that he's putting into these issues. Not because he claims to have all the answers, none of us do, but because he clearly understands the responsibility weighing on their shoulders that comes at this moment in history and time, he's treating the prospect of AI surpassing human intelligence. Not as some victory lap or pot of gold, but as a moral test for sure, we're gonna hit bumps.

Some of them would be tech, some of them will be technical. Some of them will be political, some will be economic, some will be self-inflicted and some, there's nothing we could do about. But reading these essays gave gave me something that I don't after.

I don't feel after reading or hearing about AI thought leaders, it gave me hope and hope's a precious thing. Hope that there are adult, there are indeed adults in the room. Hope that at least some of the people raising this incredibly powerful adolescent teenager that we have before us, understand what the stakes really are here.

Hope that intelligence, humility, and responsibility might just scale alongside capability. AI is growing up fast. The real question is not whether it will mature.

The real question is whether we will mature and that my friends, is what Shimmy says. I'll see you next week. By the way, I wrote a two part article on this.

You can check it out on text drawing ai. There's a lot in there to the digest. And if you're interested in the court constitution, I have another article coming out tomorrow, Friday, just on the Constitution and check that out.

But until then, man, I'll see you next week on Shimmy says, have a great weekend everyone.